The Poet in the Age of Artificial Intelligence by Matthew Leger
In the coming decades, artificial intelligence will change the way we view, consume, and create
poetry. Much of the debate concerning the role AI may play in the future of art is framed in a
way that places the two at odds – the situation is, however, far more nuanced than this binary
implies. Poets willing to embrace technological change will have a part in creating a new poetry,
one formed symbiotically with machine.
In the 1967 poem, “All Watched Over by Machines of Loving Grace,” Richard Brautigan muses:
I like to think (and
the sooner the better!)
of a cybernetic meadow
where mammals and computers
live together in mutually
programming harmony
like pure water
touching clear sky.
This poem presents an unusually positive portrait of a future in which human life is deeply intertwined with artificial intelligence. In other media — especially cinema — depictions of AI are generally bleaker, often showcasing devastated worlds where desperate humans, usurped of their role as cosplay gods, have been pitted against their robotic creations. At their core, these depictions rework the moral at the heart of countless myths: the story of humans going too far. Yet as AI takes on an increasingly large role in our day-to-day lives, it remains to be seen whether such warnings will continue to seem necessary. On a less mythical level, headlines echo media pessimism, and no wonder: as factory workers around the world lose jobs to automation, talk about the ramifications of AI has spread across occupations, including now even the professions. Still, we hear relatively little discourse around AI in the arts. Maybe this is because art, linked so closely with the human condition, is thought to be intrinsic to us, or even sacred, and thus inimical to “artificial” intelligence; maybe it’s a result of plain age-old hubris; or perhaps AI, still in its infancy, simply isn’t worth worrying about yet for most, even if neural networks are learning more by the day, expanding like vine across binary planes.
As a young, aspiring poet, the superfluity of futures shimmering behind the horizon often weigh heavily on me. Like any craft, poetry requires a great deal of investment in both time and spirit; the idea of devoting a lifetime to it, only to be relegated to the dustbin in favor of infinitely more capable machines is, understandably, anxiety provoking. My research has shown me that this is an unlikely prospect. The truth is, at the moment, artificial intelligence is nowhere near powerful enough to take the poet’s place. To be sure, it’s hard to predict what will happen over the course of our lifetimes; though it’s unlikely, we certainly can’t rule out the possibility that artificial intelligences may be able to rival us as poets in the coming decades. Still, to simplify artificial intelligence in poetry to such a question misses the bigger picture — for the poet willing to embrace technology as it develops, there’s a great deal to be optimistic about. Beyond the pure aesthetic value of an art that’s both sprung from and foreign to us, it’s quite possible the future of poetry resides in a symbiotic relationship between man and machine. And perhaps in some ways the emergence of that future will reveal such symbiosis to have characterized poetry in the past. Moreover, as questions surrounding sentience become increasingly consequential, poets — as purveyors of a craft lauded for its expression of subjectivity — are in a unique position to help determine the bar marking what counts as true intelligence.
Programming the Muse
Language is, by many accounts, a tool of liberation, the wellspring from which we draw in order to express ourselves; simultaneously, as the structure through which we organize our thoughts, we are confined within its ever-growing walls. Although the poet seeks to subvert language structure’s tyrannical grip on meaning, it’s undeniable that language conditions us to think in certain ways, patterns that a computer remains unconstrained by. As technology continues to advance at a rapid pace, the poets of the future may very well be the poets willing to open themselves to the machine and form a symbiotic creative relationship, allowing for ameaningful, aleatory poetry.
The surrender of complete intellectual control in pursuit of a greater art isn’t unprecedented — the French and Spanish Surrealists of the 20th century used a variety of methods to create a more entropic poetry. In Robert Bly’s book, Leaping Poetry[1], Bly translates a beautiful passage by Federico García Lorca, writing:
Very often intellect is poetry’s enemy because it is too much given to imitation, because it lifts the poet to a throne of sharp edges and makes him oblivious to the fact that he may suddenly be devoured by ants, or a great arsenic lobster may fall on his head. (Bly, 41)
Lorca’s point — which itself is enacted towards the end of his sentence — is that, in limiting ourselves to a strictly coherent poetry, we risk losing the pleasures and potential lessons intuited from a poetry that utilizes wild association. Bly uses the term “leaping” to describe the fantastic moments in poetry in which something that appears to be nonsensical on the surface carries a powerful, evocative charge, one that is both satisfying and stimulating. This “leap,” he writes, “can be described as a leap from the conscious to the unconscious and back again, a leap from the known part of the mind to the unknown and back to the known” (Bly, 1). While poets like Lorca toiled away matching disparate combinations of words and phrases to achieve fantastic leaps, computer savvy poets now have artificial intelligence at their disposal, allowing them to maximize their time and experiment with a broader range of material to achieve unique leaps of association.
The writings of André Breton and the French Surrealists are also of particular value to the machine-oriented poet. In Manifestoes of Surrealism[2], Breton advocated for what he called “automatic writing,” a method in which the writer allows unconscious thoughts to spill onto the page, uninhibited. Breton explains:
It is, as it were, from the fortuitous juxtaposition of the two terms that a particular light has sprung, the light of the image, to which we are infinitely sensitive. The value of the image depends upon the beauty of the spark obtained; it is, consequently, a function of the difference of potential between the two conductors . . . Now, it is not within man’s power, so far as I can tell, to effect the juxtaposition of two realities so far apart. The principle of the association of ideas, such as we conceive of it, militates against it. (Breton, 37)
Though Breton sees automatic writing as a method that’s distinct from regular writing, he does not believe it’s artistically illegitimate or antagonistic; he sees automatic writing as a component central to the future of poetry. Breton’s use of mechanical imagery is especially prescient, as the sentiment he expresses holds as well for computer-generated poetry. By carefully selecting source material to input into a generator, the poet can bypass the conscious associations Breton so deplores and excavate for meaning amidst the resulting disjunctions. Moreover, if the poet uses Breton’s method of automatic writing for their source material, the generator’s results may be even more spectacular. Perhaps the most directly applicable practice Breton advocates for is the newspaper cut up method; of this, he writes:
Surrealist methods would, moreover, demand to be heard. Everything is valid when it comes to obtaining the desired suddenness from certain associations . . . It is even permissible to entitle POEM what we get from the most random assemblage possible (observe, if you will, the syntax) of headlines and scraps of headlines cut out of the newspapers. (Breton, 37)
It is in both method and philosophy that Breton’s work exists as a sort of spiritual precursor to computer-generated poetry. In a similar vein, the “cut up method,” in which texts are cut up and rearranged into new texts, has also maintained an enduring popularity among writers. In the broader scope of literary history, then, computer-generated poetry represents a logical progression of these methods, as well as a major deviation in scope.
The Significance of Poetry in Testing AI Capacity
Alan Turing’s essay “Computing Machinery and Intelligence” is best known for how Turing outlines his famous thought experiment, The Imitation Game, which proposes a way in which artificial intelligence can be judged as sentient. Less well known, is the second thought experiment Turing proposes, in which an AI is asked to “write [the judge] a sonnet on the subject of the Forth Bridge”—to which request the computer replies “Count me out on this one. I never could write poetry” (Turing, 2). Turing’s inclusion of poetry in the discussion of computing intelligence is meaningful. In Kurt Beals’ essay, “‘Do the New Poets Think? It’s Possible’: Computer Poetry and Cyborg Subjectivity, Beals argues that “The frequent characterization of poetry as a medium for the expression of subjectivity and interiority made computer poetry an ideal testing ground for new and different modes of subjectivity.”[3] Just as a schoolteacher struggles to convey the complexity of T.S. Eliot’s “The Love Song of J. Alfred Prufrock” to a room of aloof 11th graders — resorting to literary devices and historical context in order to help the students feel the text — programmers feed poems, metrical data, and grammatical rules to algorithms to help them process text in ways that will matter to society. But the question remains: at what point, if any, is a machine-generated poem legitimate? After all, computational poetry generators are advanced enough to follow grammatical rules and adhere to structural constraints such as meter and rhyme. Further complicating the question is that — as proven by the surrealist’s automatic writing — evocative writing needn’t always be coherent, just as the coherency of a poem doesn’t ensure it’ll be evocative. According to Beals, Turing argues “that the subjective experience of thought and emotion cannot be invoked as the decisive factor in distinguishing human from artificial intelligence. As long as the computer is able to generate poems (as well as commentaries or explications) sufficiently similar to those of human authors, it has passed Turing’s poetry test—there is no need for it to feel the emotions that those poems purport to express.” (Beals, 157) If we no longer exist in an age in which great poetic achievements are believed to have been bestowed to us by the divine, then a poem — like the poet — is a product of the specific set of material conditions that created and prompted the poet to write it.
This same model is however amenable to an opposed view, on which if machines are products of their own material conditions, then the programmer could be viewed as the computer’s muse, instilling it with language drawn from an incomprehensible source. If we’re willing to accept machine-generated poetry as at once both an inevitable manifestation of language development and a form of art that is alien and potentially beautiful, we open ourselves to new poetic horizons. In his New Yorker article “What Happens When Machines Learn to Write Poetry,” Dan Rockmore makes an eloquent case for machine-generated poetry:
Given the power of new techniques in artificial intelligence, why not think more broadly about the kinds of art one can make using it? We could think of “machine-generated” as a kind of literary G.M.O. tag—or we could think of it as an entirely new, and worthy, category of art. As we interact more and more with machines, both knowingly and unknowingly, our own expectations around both work and art will change, and labels will start to dissolve.
Although answers to the question of what makes us human are sure to be varied, many would agree that we are defined by our use of complex language, our creativity, and our self-awareness. Developments in artificial intelligence have added a new layer of complexity to this discourse; like any major scientific development, AI will undoubtedly change the way we view ourselves, and, by extension, the way we view subjectivity and creativity. As Rockmore says, in time, labels will dissolve; and should we choose to embrace computer-generated poetry sooner rather than later, we will attain a better understanding of ourselves and our art. Still, to contextualize machine-generated poetry in the present, it helps to imagine what a truly autonomous, non-human poetry could look like in the future.
The Author, Vanished
It’s not easy to envision a non-human poetry; while Romantics such as John Keats — “The Poetry of earth is never dead” — may claim rivers are ellipses or canyons enjambments, for our purposes, we’ll limit ourselves to the written word.[4] In Italo Calvino’s remarkably prescient essay, “Cybernetics and Ghosts,” Calvino envisions what a truly mechanical, autonomous work of poetry could look like.[5]Calvino is careful to define this “true literature machine” as one that could “bring to the page all those things that we are accustomed to consider as the most jealously guarded attributes of our psychological life, of our daily experience, our unpredictable changes of mood and inner elations, despairs and moments of illumination” (Calvino, 16). Thus far, we’ve considered the implications of a symbiotic relationship between poet and machine; Calvino, imaginative as ever, is interested in a machine that appears to be able to think and feel, or is at least complex enough to mimic human thought to the point of inseparability — in other words, complex enough to pass the Turing test.
Calvino believes that such a machine would, firstly, be capable of producing classical works, works requiring set meter and rhyme schemes. But Calvino adds an important caveat:
The true literature machine will be one that itself feels the need to produce disorder, as a reaction against its preceding production of order: a machine that will produce avant-garde work to free its circuits when they are choked by too long a production of classicism. In fact, given that developments in cybernetics lean toward machines capable of learning, of changing their own programs, of developing their own sensibilities and their own needs, nothing prevents us from foreseeing a literature machine that at a certain point feels unsatisfied with its own traditionalism and starts to propose new ways of writing, turning its own codes completely upside down. (Calvino, 16-17)
This essay was written in 1967, and technology has made great strides since; the catch, however, is the phrase “feels unsatisfied,” which indicates that a machine would have to feel — would have to make a conscious decision — in order to revolt against the constraints of traditionalism. That said, we must keep in mind how, as Beals outlines when he develops Turing, “the subjective experience of thought and emotion cannot be invoked as the decisive factor in distinguishing human from artificial intelligence”: how “there is no need for [the computer] to feel the emotions that [its] poems purport to express” (Beals, 157). And since we have no way to know if a computer is actually “feeling” or not, if Calvino’s metricof reaction and revolt is to provide a more valuable way to evaluate computer-generated poetry, it will have to do so without such response being thought contingent on satisfaction. Calvino notes that, though there are a plethora of theories out there that try and pinpoint where inspiration comes from — be it from the divine, chemical fluctuations in the body, true genius, or wherever — nobody can say for certain just how, when Milton sat down to write his masterpiece, he was able to do it. Accordingly, we can’t say for certain that a machine will never be able to write great poetry.
One thing we can do is analyze the history of literature to see how new traditions have come to form. To this end, Calvino writes that
Literature as I knew it was a constant series of attempts to make one word stay put after another by following certain definite rules; or, more often, rules that were neither definite nor definable, but that might be extracted from a series of examples, or rules made up for the occasion—that is to say, derived from the rules followed by other writers. (Calvino, 18-19)
If machines are judged on their ability to meld preexisting literary traditions into new traditions, in the same way new poets break the rules of their predecessors, it seems to me this would constitute a genuinely alien, autonomous poetry. Moreover, if machines are eventually able to craft a new sort of literary tradition, it would be an indescribably fruitful development for human poets: we would have access to a vast collection of material to absorb and, in turn, revolt from, establishing a dialectical relationship that could prove mutually beneficial to man and machine. After all, poetry has helped us understand ourselves better; if a machine were to develop a capacity for consciousness, understanding poetry would be a good place to start.
Ultimately, such speculation on consciousness is not germane to projecting the most realistic future of poetry as being one in which a dialectical relationship between man and machine helps to craft a new, enriching poetic tradition, just as the formalization of the typewriter created new possibilities with line spacing and enjambement. Calvino believes that this new relationship will cause mankind to fundamentally alter their view of the author:
What will vanish is the figure of the author, that personage to whom we persist in attributing functions that do not belong to him . . . And so the author vanishes — that spoiled child of ignorance — to give place to a more thoughtful person, a person who will know that the author is a machine, and will know how this machine works. (Calvino, 20)
Although it’s true that, at the moment, the author exists as a largely opaque figure in our culture, recent developments in AI have begun to attract increased attention from writers. In June of 2020, Open AI released the beta for their new GPT-3 language model; according to Wired.
GPT-3 was built by directing machine-learning algorithms to study the statistical patterns in almost a trillion words collected from the web and digitized books. The system memorized the forms of countless genres and situations, from C++ tutorials to sports writing. It uses its digest of that immense corpus to respond to a text prompt by generating new text with similar statistical patterns.
The reaction to GPT-3 among programmers – and computer-savvy writers – has been impressive; while the program has been championed for its capacity to generate astoundingly fluid texts, critics point out that many of its results lack coherency. This argument regarding coherency lies at the crux of the debate surrounding computer-generated poetry – how can one derive personal meaning from a poem if they know it to be the result of a randomly generated process? To this I say we must consider how we glean meaning from any poetry in the contemporary moment.
I’ve found that the best poetry I’ve written is serendipitous, resulting from the unforeseen placement of two lines next to each other, experiments with enjambment that recontextualize the preceding line, and things of that nature. No writer can deny the joy they feel when a wild theory about their poem is thrown out in workshop; good writers may even adopt the occasional theory and run with it. This is undoubtedly one of the greatest things poetry has to offer the writer and the reader — bait to dip into the rivers of the unconscious, a centrifuge in which complex feelings and memories can be inserted, deconstructed, and laid bare. Poetry generated from a machine is a poetry removed from ourselves, a perspective exterior to us, a sublimate of the serendipity that makes writing an intoxicating process. Insofar as poetry is a vehicle for social change, a machine-generated poetry could help expose the conscious and unconscious biases inherent to literary traditions, the blind spots that perpetuate systems of oppression. While there’s no way to remove the literature we’ve been raised on from our heads, programming artificial intelligence with large amounts of carefully curated source material — a diverse array of texts that are more representative of historically oppressed groups than literary traditions account for — would result in a richer poetry of the future, one that is welcoming to all poets for experimentation. Calvino offers a valuable analysis of the gulf between computer generation and human interpretation of a text when he notes that “the literature machine can perform all the permutations possible on a given material, but the poetic result will be the particular effect of one of these permutations on a man endowed with a consciousness and an unconscious, that is, an empirical and historical man. It will be the shock that occurs only if the writing machine is surrounded by the hidden ghosts of the individual and of his society” (Calvino, 26-27). Were a poet to process a comprehensive anthology of a specific literary canon through a powerful computer generator, the resulting poems would paint an enlightening portrait of the values, customs, and prevailing ideological tenets of the culture that produced the canon. In other words, the machine, shielded from the hidden ghosts of a society, is uniquely qualified to produce infinite quantities of mirrors for our gazing pleasure, resulting in a Baroque Palace of the Collective Unconscious. It is then the poet’s duty to fill the space in between lines with ghosts.
If we are to build powerful artificial intelligence someday, we must make sure it’s capable of producing poems that are representative of all voices in our diverse world. This is a process that must occur while artificial intelligence develops, as those who’ve studied literature know that catch up work is all too difficult. The release of GPT-3 has provided us with an early, all-too-enlightening example of the importance of careful curation. Wired reports:
Facebook’s head of AI accused the service of being “unsafe” and tweeted screenshots from a website that generates tweets using GPT-3 that suggested the system associates Jews with a love of money and women with a poor sense of direction. The incident echoed some of WIRED’s earlier experiments in which the model mimicked patterns from darker corners of the internet.
Here we come to understand the importance of the computer-literate poet: If humans are to function as the muse of the first great machine, we must endow this machine with a poetic sensibility, so as to chart an ethical course for it to grow along and prevent it from aiding and abetting those who seek to make the world a worse place. While there may be a spot at the table for poets in crafting a technological future, it’s a moral imperative to fill it – doing so will honor the disregarded voices of history and leave us with a more illuminating portrait of ourselves.
The Static, Humming Future
There’s no telling where artificial intelligence will take us over the remainder of the century. While the resonances of change are sure to be vast – one can already perceive the faint prowl of progress in university halls and research institutes – the machine-oriented poet, ears attuned to the subtle hum of static, finds himself at the precipice of a new poetry, a poetry grown out of himself. In my time utilizing poetry generators, I’ve become hyperaware of the seemingly innate, human tendency to search for meaning in everything – though I know how the generator’s algorithm works, every so often I’m fed such a finely sequenced set of words I can’t help but wonder if there’s something greater at work behind my screen. But the truth is, this sort of speculation is immaterial; for, in imparting language to machine, the poet becomes the muse; he is permitted a glance into whatever it is that prompted mankind so very long ago to begin writing poetry in the first place. To embrace the alien beauty of verse crafted by computers is to shed light on what makes human verse unique, resulting in a more sonorous, enlightening poetry for all.
****************************************************************************
My Computer-Generated Poems
For my first experiment with a poetry generator, I drew inspiration from Carl Dennis’ poem “Two Lives,” and decided to write a poem imagining an alternate version of my life, as — given the way our lives are dictated by sheer. blithe chance — it provides a good template for this type of poetic exploration.
I used the free jGnoetry generator, which allows you to create a template of syllables, input sections of text into corpuses, and assign the corpuses a weight, dictating how often they’re to be used — the text supplied functions as the generator’s vocabulary, which it uses to construct a poem in the form of your choosing. Once your poem is generated, you can select which words you want to keep, and the program generates new words around them. My intent in writing this poem was to create three stanzas centered on different, impactful moments of my life; by inserting them into the generator, I wanted to see what these three distinct moments would look like melded together. Here is the original poem:
In my other life
I am tossing a peach bone
off my grandmother’s veranda
half-listening to her Delphic whispers.
She quit smoking 40 years back
so oxygen machines remain as foreign
to us as Kentucky graveyards.
In this life I have never seen my father
weep, or excavated God in a Motel room.
I brood over guilt with remedy.
In this life there is more death to come
it will be as incomprehensible as ever
or maybe it won’t be
my grandmother tells me. I believe her
pick apples before they fall
a silent mocking gesture to grief.
In this life everything can be explained
at least for a few more years.
In my other life
I am walking inside with dirty knees.
From my window, I watch the neighbor boy
hang stag antlers on his wall, kiss his guns,
and dip his hands in bliss. Due to a clerical
error, I store my helmet in a corner opposite
the skunk-musk smell of rolling papers
in the infinite expanse of the 8th grade
locker room. In this life I will never meet
the lanky red-haired boy who could show me
how to fix myself for a couple of hours.
I will not come to know the locking sound
of medicine cabinets; my hymns will remain
clean, pills will not wink in newspapers.
In my other life I will hear of a funeral
for my classmate, feel a passing grief,
and forget. There will be no whiskey
to pour on a grave not too far from my house.
I am rinsing my dirty knees; the future is bright
perhaps I will go hunting next week and feel
nothing for the pathetic animals. I hug my
shadow before bed, hope he returns by morning.
In my other life
I am biking up the Georgia cul-de-sac
to the ramshackle park lost in trees.
On the swing-set, my body spills
out of my body; flowers nearby sprout
with every thrust, silence outpaced and
stowed away for later use. Gazing into
the horizon, Texas is as distant as a dream.
In this life, I head with Tracey to the pond
where tadpoles wade as if hatched
in warm milk; we squeeze until summer
oozes out of their throats, watch them pop
like poppy pods, dip our hands in mud.
I promise that I will stay here forever.
I am not wrong. In this life me and Tracey nap,
awake heirs to evening, recipients of a stillness
so ripe it has no choice but to breathe itself
into silt once we leave.
One of the benefits of using a generator is to create order out of disorder. For my generated poem, I settled on five quatrains, the first line of each stanza five syllables (the amount of syllables in the opening phrase, “In my other life”) and the next three lines ten syllables, as pictured below.
Next, I created three corpuses, one for each stanza, and weighted them.
I adjusted the weight of each corpus throughout the generating process. I was really happy with my final results:
Poem #1: In My Other Life
In my other life
there is more to the future. In a motel
room she quit smoking the silence I believe
her; there gazing into my other life
I am walking to
her funeral in newspapers. I am
rinsing my shadow before bed, my hymns
will sprout oxygen machines; my summer
oozes out of rolling
papers that I am tossing, the pills so
ripe flowers nearby sprout. Georgia cul-de-sac
dirty like Texas in mud. I store throats,
or maybe smell of
medicine cabinets; remain clean, hope
returns here: this red-haired boy with dirty
knees. In this life I have to fix myself
for a peach bone hug.
Gazing to grief, or maybe due to evening,
I excavated god in mud. In my
Grandmother, distant machines remain clean.
* * * * * * * * *
Besides some punctuation I added and changed, the poem is completely unedited.
Next, I inputted an old poem of mine, titled “Bees,”; though I liked it, I gave up on it, having arrived at an impasse. Here’s the original poem:
Bees
I.
Lullabies are sweetest
at the bottom of the sea
I felt the mothers say,
but it’s nothing like
the ocean’s pull
from room 472
at the Shearwater Inn,
where a faceless woman
is shoving hypodermic bees
into my fingertips. Somebody
is knocking on the door.
We have been married
for a week
to this room
vomiting off the balcony,
sunbathing in the washroom,
subsisting on the loose screws
of our stopgap heaven.
When my fingers go numb,
I rewrite our meeting: this time,
I watch your steady breath
buoying magnolia petals
& once you open your wallet
it is as if miles of light slip
out like angel pus; I see
your smile for the first time
& it is suddenly unfathomable
to think of your son’s picture
on a cheap motel floor, his face
sagging below a crown of pubes,
unfathomable to think a recently
ordained jukebox allowed us this
island of crusty towels & expired
creams. With every knock we are
more unfathomable, more revolting,
until we are forced to cede our eyes
to the quiet humming of bees:
a sting
You lick honeydew from the corner
of my lips & I am okay.
II.
The people of the Shearwater Inn
nod behind cheap balustrades like
deposed kings, eating credit cards
like slop, eyes fixed on the coast.
There is knocking,
always knocking, sirens
prowling leisurely,
floaties popping.
On the beach,
a stroller rolls
closer everyday
but I can only move
in sways.
III.
nightstand cradling
an upside-down picture,
we are vibrating:
it is no longer a knock,
but a thud—
i imagine the sea quaking
but i am small
& my lover
limp
i don’t have the guts
to check
IV.
I am watching an adoption,
a lone stroller by the beach;
hundreds of fizzled mothers
spar with crayons
over flushed cheeks
* * * * * * * * *
After separating the stanzas into different corpuses, I added an extra corpus composed of nonsensical phrases, in an attempt to use the French Surrealist’s automatic writing method:
Here’s the final result:
Poem #2: Bees
Somebody is suddenly unfathomable
To check I imagine
Eyes to think of light
Popping on the beach like
A knock eating cheap balustrades
Aging sea-light beckons
The coast there lucid, relishing
When my lips, dirt-caked &
Open think am i okay
We are deposed credit cards
The quiet sunbathing graveyard allowed us
This slurred lifetime
Shoving hypodermic bees, expired
Creams with every nod
magnolia petals behind cheeks & lakes
vibrating in our fingertips
I imagine lullabies
Strummed upside-down
A stroller in styrofoam pulls a horse
Unfathomable there the beach foaming limp
Floaties popping on the ocean angel pus
Cradling light in sways, mothers rise ordained
Dehydrated balloons with faux-knot necks
* * * * * * * *
Poem #3: Dream Song
For this poem, I set out to write my own Dream Song. I fed the generator my favorite poems from John Berryman’s The Dream Songs[6] – #1, #3, #4, #8, #14, #25, #29, #55, and #76 – and wrote three stanzas of sestets, as Berryman generally did. The following poem is composed entirely of words from Berryman’s writing:
Black sobriety sober as mirrors gone.
Sound is boring spelled of being —if
—I’m not —I cannot remember I conclude
man heals sideways —a glorious
sloshed psychiatrist hungers
Mr ‘underneath betrayed to give me low tide
I mentioned he tamped euphoria &
somehow health leaves
a pity oh well
my installed all-disappointed bafflin
odd I admit you have come out a blur
a huffy glorious face renders
hope all these things but more & smoother
growing blind and sloshed like a hundred
years Mr Bones pled sooner
heaven & helpful myrrh
what elongates bores him the sense there
is nothing— a jerk I join Rilke great & valved.
* * * * * * * * *
Poem #4: Spring, 2021
The mouth of far-off flowers grinds its teeth:
for the first time I hear the call of spring.
I approach in the breeze — heaven wakes,
seemingly for me, I offer the blood-splatter
of my knuckles, play peek-a-boo
with a violin in the ears of the greenery.
For a moment all is laminated;
I will no longer wage war on fluttering.
* * * * * * * * *
Poem #5: The Girl Who is Like a Pear
To see what I could get inputting a multitude of poet’s work into the generator, I inputted three poems from three of my favorite writers: Denis Johnson’s “The White Fires of Venus”; Larry Levis’ “Poem Ending with a Hotel on Fire”; and James Tate’s “The Wheelchair Butterfly.” My experience constructing this poem was unique; it felt like an appropriate homage to my idols:
They looked beyond the face, where they are not!
Change shows all a moment;
yesterday was warm, bartender dead!
I know what underground heart is urinating
your extinction & for carfare that chirring,
where the ordinary hornets
in this world sleep on into death, I thought
How I’m telling you, O warden of fingerprints,
us inhaling gasoline like a ravine . . .
The girl who is like a pear
rides her blouse & in this is dying.
The soul curls up
by ambulance sirens
a dandelion sends
tossed instructions
to a fountain where a butterfly
froze; the overcast had to do with it
Phonily you could hear the initiation tape
where shadow figures drawn out the head
swirl removed then waking reeling bruised
I thought they looked emaciated
past the plains whispering daydreams,
distant amusement park music, beware
the calm mayor
O how the trumpet once meant
to ring the underground
Closed and lulled a human’s heart
may become gradually confused with
America: they own your face, beware your
hair — a glass of fleas — & the history book,
the way they bicycle backwards to touch things of
yesterday beware the seeds where they say the
contorted suicide waits once an alarm blares
The soul curls up
by ambulance sirens
The soul curls up
by ambulance sirens
* * * * * * * * *
[1] Bly, Robert. Leaping Poetry. University of Pittsburgh Press, 2008.
[2] Breton, André. Manifestoes of Surrealism. First, University of Michigan Press, 1969.
[3] Beals, Kurt. “‘Do the New Poets Think? It’s Possible’: Computer Poetry and Cyborg Subjectivity.” Configurations (Baltimore, Md.) 26.2 (2018): 149–177, 154. Web.
[4] Keats, John. “On the Grasshopper and Cricket by John Keats | Poetry Foundation.” Poetry Foundation, https://www.poetryfoundation.org/poems/53210/on-the-grasshopper-and-cricket. Accessed 17 Aug. 2020.
[5] Calvino, Italo. The Uses of Literature: Essays. Translated by Patrick Creagh, Houghton Mifflin Harcourt, 2017.
[6] Berryman, John. The Dream Songs. Farrar, Straus and Giroux, 2014.
Matthew Leger is an August 2020 graduate of the University of Texas at Austin, where he studied English and Creative Writing. An aspiring poet, and recent winner of the 2020 Andrew Julius Gutow Prize, his work can be found here. When he’s not writing, he’s likely to be found holed away in his room recording music, courting the ire of his neighbors at strange hours of the night. He plans to pursue an MFA in Fall of 2021.
